Paragraph opener scorer

Vocabulary is easy to fix and easy to fake. Structure is harder. A generated draft tends to announce its own paragraphs: Additionally opens the second point, Moreover the third, Ultimately the conclusion. The words are correct and the paragraph is empty, because the connective is doing the job the argument should do.

Measuring it per paragraph rather than per document gives you something actionable. You are not told that the piece has eight connectives; you are shown which paragraphs start with one, and the fix is always the same: delete the opener and start on the subject.

The paragraph is a real node in a Lexical document, so the marker can be attached to it as you type rather than recomputed for a preview. That means the flag survives editing: rewrite the opening words and the tint lifts on the keystroke that changes them.

It also means the tool works the way the writing does. You are not pasting, running and reading a report; you are writing, and the paragraph tells you it opened badly while you can still remember what you meant.

The percentage is a reading aid, not a grade. Human writers open with connectives too, and occasionally a paragraph genuinely wants one. Above about one in five, the scaffolding becomes the voice, and that is worth a rewrite regardless of who or what produced the text.